Abstract
The odds ratio and the relative risk are commonly used as measures of association in a fourfold table. The bias of the standard point estimators of these parameters is investigated for small samples. Simple alternative estimators that possess superior performance in terms of bias and mean squared error are suggested.